General Question Answering on PopQA (EM, FBERT)
47.87EMDynamic Collaboration Framework
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Dynamic Collaboration FrameworkSLM=Qwen3-8B2026.04 | 47.87 | 75.58 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.1-8B2026.04 | 45.54 | 74.43 | |
| CoTSLM=Qwen3-235B2026.04 | 45.44 | 73.08 | |
| Dynamic Collaboration FrameworkSLM=Qwen3-4B2026.04 | 44.17 | 74.08 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.2-3B2026.04 | 41.5 | 72.17 | |
| Direct InferSLM=Qwen3-235B2026.04 | 38.38 | 67.64 | |
| CoTSLM=Llama-3.1-8B2026.04 | 25.59 | 61.45 | |
| CoTSLM=Qwen3-8B2026.04 | 20.69 | 58.95 | |
| CoTSLM=Qwen3-4B2026.04 | 16.04 | 57.28 | |
| CoTSLM=Llama-3.2-3B2026.04 | 14.22 | 52.65 |